% Table 3
% November 4, 2015
%


clear all;
small = 1.0e-10;
big = 1.0e+6;                  
rng(63761);
 
  % -- File Directories  
  outdir = '/Users/mwatson/Dropbox/p_comp/revision/matlab/out/';
  figdir = '/Users/mwatson/Dropbox/p_comp/revision/matlab/fig/';
  matdir = '/Users/mwatson/Dropbox/p_comp/revision/matlab/mat/';    
  
  % -- Read in Data --- 
  load_data = 1;  % 1 if reloading data from Excel, etc 
  mtoq_agg = 3;   % Temporal aggregation indicator of monthly to quarterly data
  pcomp_data_calendar_m_and_q;
  % Data Series Used
  dp_agg = dp_agg_q;
  dp_agg_xfe = dp_agg_xfe_q;
  dp_agg_xe = dp_agg_xe_q;
  dp_disagg = dp_disagg_q;
  share_avg = share_avg_q;
  calvec = calvec_q;
  dnobs = dnobs_q;
  calds = calds_q;
  nper = 4;
  label_suffix = '_Q3';
  
  % Dates
  first_date = [1960 1];
  last_date = calds(end,:);
  ismpl_60_end = smpl(calvec,first_date,last_date,nper);
  cal_60_end = calvec(ismpl_60_end==1);
  calds_60_end = calds(ismpl_60_end==1,:);
  dnobs_60_end = size(cal_60_end,1);
  
  % First POOS Date
  first_date_poos = [1990 1];
  tmp = calds_60_end == repmat(first_date_poos,dnobs_60_end,1);
  [a,i_first_poos] = max(sum(tmp,2));  % Index in calvec_ismpl of first poos period
  
  % Date for 2008:Q4 -- skip
  if nper == 4;
   date_skip = [2008 4];
   tmp = calds_60_end == repmat(date_skip,dnobs_60_end,1);
   [a,i_skip] = max(sum(tmp,2));  % Index in calvec_ismpl of first poos period
  else; 
   date_skip = [2008 10];
   tmp = calds_60_end == repmat(date_skip,dnobs_60_end,1);
   [a,ii] = max(sum(tmp,2));  % Index in calvec_ismpl of first poos period
   i_skip = [ii ii+1 ii+2]';
  end;
  
  outfile_name = [outdir 'fcst_rslt_poos' label_suffix '.out'];
  fileID = fopen(outfile_name,'w');
  
  dp_agg_60_end = dp_agg(ismpl_60_end==1,:);
  dp_agg_xe_60_end = dp_agg_xe(ismpl_60_end==1,:);
  dp_agg_xfe_60_end = dp_agg_xfe(ismpl_60_end==1,:);
  dp_disagg_60_end = dp_disagg(ismpl_60_end==1,:);
  share_avg_60_end = share_avg(ismpl_60_end==1,:);
  
  dpa_agg = zeros(dnobs,1);
  dpa_agg_xe = zeros(dnobs,1);
  dpa_agg_xfe = zeros(dnobs,1);
  dpa_disagg =  zeros(dnobs,n_incl);
  for i = 1:nper;
    dpa_agg(nper:end) = dpa_agg(nper:end)+dp_agg(i:end-nper+i);
    dpa_agg_xe(nper:end) = dpa_agg_xe(nper:end)+dp_agg_xe(i:end-nper+i);
    dpa_agg_xfe(nper:end) = dpa_agg_xfe(nper:end)+dp_agg_xfe(i:end-nper+i);
    dpa_disagg(nper:end,:) = dpa_disagg(nper:end,:)+dp_disagg(i:end-nper+i,:);
  end;
  dpa_agg = dpa_agg/nper;
  dpa_agg_xe = dpa_agg_xe/nper;
  dpa_agg_xfe = dpa_agg_xfe/nper;
  dpa_disagg = dpa_disagg/nper;

  dpa_agg_60_end = dpa_agg(ismpl_60_end==1);
  dpa_agg_xe_60_end = dpa_agg_xe(ismpl_60_end==1);
  dpa_agg_xfe_60_end = dpa_agg_xfe(ismpl_60_end==1);
  dpa_disagg_60_end = dpa_disagg(ismpl_60_end==1,:);
  
  % -- Univariate Filtered and Smoothed Values
  ulabel = ['dp_agg_poos' label_suffix];
  str_tmp = [matdir 'tau_mean_pct' ulabel]; load(str_tmp);
  tau_f_h = tau_mean_pct(:,1);
  tau_f_mean_pct_h = tau_mean_pct;
  
  ulabel = ['dp_xe_poos' label_suffix];
  str_tmp = [matdir 'tau_mean_pct' ulabel]; load(str_tmp);
  tau_f_xe = tau_mean_pct(:,1);
  tau_f_mean_pct_xe = tau_mean_pct;
  
  ulabel = ['dp_xfe_poos' label_suffix];
  str_tmp = [matdir 'tau_mean_pct' ulabel]; load(str_tmp);
  tau_f_xfe = tau_mean_pct(:,1);
  tau_f_mean_pct_xfe = tau_mean_pct;
  
  % -- Some results for the 17-component model
  mlabel = ['_mucsv_tvp_17c' label_suffix];
  str_tmp = [matdir 'agg_tau_mean_pct_poos' mlabel]; load(str_tmp);
  tau_f_17c = agg_tau_mean_pct(:,1);
  tau_f_mean_pct_17c = agg_tau_mean_pct;
  mlabel = ['_mucsv_constAlpha_17c' label_suffix];
  str_tmp = [matdir 'agg_tau_mean_pct_poos' mlabel]; load(str_tmp);
  tau_f_constAlpha_17c = agg_tau_mean_pct(:,1);
  tau_f_constAlpha_mean_pct_17c = agg_tau_mean_pct;
%  tau_f_17c = tau_f_h;
  
  % -- Some results for the 3-component model
  mlabel = ['_mucsv_tvp_3c' label_suffix];
  str_tmp = [matdir 'agg_tau_mean_pct_poos' mlabel]; load(str_tmp);
  tau_f_3c = agg_tau_mean_pct(:,1);
  tau_f_mean_pct_3c = agg_tau_mean_pct;
  mlabel = ['_mucsv_constAlpha_3c' label_suffix];
  str_tmp = [matdir 'agg_tau_mean_pct_poos' mlabel]; load(str_tmp);
  tau_f_constAlpha_3c = agg_tau_mean_pct(:,1);
  tau_f_constAlpha_mean_pct_3c = agg_tau_mean_pct;
%  tau_f_3c = tau_f_h;
  
  nhorizon_years = [1 2 3];
  for ih = 1: size(nhorizon_years,2);
   
    nhorizon_avg = nhorizon_years(ih)*nper;
    % Compute Target series to forecast
    dp_avg_target = NaN(dnobs_60_end,1);
    dp_avg_target_x = NaN(dnobs_60_end,1);
  
    for t = i_first_poos:dnobs_60_end-nhorizon_avg;
      dp_avg_target(t) = mean(dp_agg_60_end(t+1:t+nhorizon_avg));
     end;
     dp_x = dp_agg_60_end;
     dp_x(i_skip) = NaN;
     for t = i_first_poos:dnobs_60_end-nhorizon_avg;
       tmp = dp_x(t+1:t+nhorizon_avg);
       dp_avg_target_x(t) = mean(tmp(isnan(tmp)==0));
     end;
  
     % Compute Forecast errors
     target = [dp_avg_target dp_avg_target_x];
     e_h = repmat(tau_f_h,1,2)- target;
     e_xe = repmat(tau_f_xe,1,2)- target;
     e_xfe = repmat(tau_f_xfe,1,2)- target;
     e_17c = repmat(tau_f_17c,1,2)- target;
     e_3c = repmat(tau_f_3c,1,2)- target;
     e_constAlpha_17c = repmat(tau_f_constAlpha_17c,1,2)- target;
     e_constAlpha_3c = repmat(tau_f_constAlpha_3c,1,2)- target;
     e_l = repmat(dp_agg_60_end,1,2)- target;
     e_l_xe = repmat(dp_agg_xe_60_end,1,2)- target;
     e_l_xfe = repmat(dp_agg_xfe_60_end,1,2)- target;
     e_a = repmat(dpa_agg_60_end,1,2)- target;
     e_a_xe = repmat(dpa_agg_xe_60_end,1,2)- target;
     e_a_xfe = repmat(dpa_agg_xfe_60_end,1,2)- target;
  
     % Compute RMSE 
  
     mse = NaN(13,2);
     se_mse = NaN(13,2);
     n_mse = NaN(13,2);
     dif_mse = NaN(13,2);
     se_dif_mse = NaN(13,2);
  
     nma = nhorizon_avg+2;
     for j = 1:2;
       e = [e_17c(:,j) e_3c(:,j) e_constAlpha_17c(:,j) e_constAlpha_3c(:,j) e_h(:,j) e_xe(:,j) e_xfe(:,j) e_l(:,j) e_l_xe(:,j) e_l_xfe(:,j) e_a(:,j) e_a_xe(:,j) e_a_xfe(:,j)];
       for i = 1:size(e,2);
        tmp = e(:,i);
        tmp1 = e(:,1);
        tmp = tmp(isnan(tmp) == 0);
        tmp1 = tmp1(isnan(tmp1) == 0);
        n_mse(i,j) = size(tmp,1);
        tmp = tmp.^2;
        tmp1 = tmp1.^2;   
        [b,vb] = hacm(tmp,nma,1);
        mse(i,j) = b;
        se_mse(i,j) = sqrt(vb);
        tmpdif = tmp-tmp1;
        [b,vb] = hacm(tmpdif,nma,1);
        dif_mse(i,j) = b;
        se_dif_mse(i,j) = sqrt(vb);
       end;
     end;
     if ih == 1;
         mse_mat = NaN(size(mse,1),size(mse,2),size(nhorizon_years,2));
         se_mse_mat = NaN(size(mse,1),size(mse,2),size(nhorizon_years,2));
         dif_mse_mat = NaN(size(mse,1),size(mse,2),size(nhorizon_years,2));
         se_dif_mse_mat = NaN(size(mse,1),size(mse,2),size(nhorizon_years,2));
     end;
     mse_mat(:,:,ih) = mse;
     se_mse_mat(:,:,ih) = se_mse;
     dif_mse_mat(:,:,ih) = dif_mse;
     se_dif_mse_mat(:,:,ih) = se_dif_mse;
         
  end; 
  
  fprintf(fileID,'All Observations \n');  
  for i = 1:size(e,2);
    for ih = 1:size(nhorizon_years,2);
      fprintf(fileID,'%4.2f ',mse_mat(i,1,ih));
      fprintf(fileID,'(%4.2f),',se_mse_mat(i,1,ih));
      fprintf(fileID,'%4.2f ',dif_mse_mat(i,1,ih));
      if ih < size(nhorizon_years,2);
       fprintf(fileID,'(%4.2f),',se_dif_mse_mat(i,1,ih));
      else;
        fprintf(fileID,'(%4.2f) \n',se_dif_mse_mat(i,1,ih));
      end;
    end;   
  end;
  
  fprintf(fileID,'\n\n Excluding 2008:Q4 \n');  
  for i = 1:size(e,2);
    for ih = 1:size(nhorizon_years,2);
      fprintf(fileID,'%4.2f ',mse_mat(i,2,ih));
      fprintf(fileID,'(%4.2f),',se_mse_mat(i,2,ih));
      fprintf(fileID,'%4.2f ',dif_mse_mat(i,2,ih));
      if ih < size(nhorizon_years,2);
       fprintf(fileID,'(%4.2f),',se_dif_mse_mat(i,2,ih));
      else;
        fprintf(fileID,'(%4.2f) \n',se_dif_mse_mat(i,2,ih));
      end;
    end;   
  end;
  
  

 